benjaminstelzer
UserKeeps the engineering result visible through the process around it.
Categories
Indexed Skills (6)
scoville-code-anti-ai-slop
Goal-first guardrail for planning, changing, testing, reviewing, or removing code and engineering artifacts. Preserve observable outcome, canonical ownership, risk, validation, and honest evidence without scope drift. Not for conceptual questions unrelated to a codebase.
scoville-plan
Repository-native planning guardrail for creating, maintaining, resuming, auditing, and handing off durable project Plans, Work Items, and Decision records through direct Markdown and YAML edits only. Use when a task invokes Scoville Plan, requests repository-owned implementation planning or decision records, must survive interruption or compaction, or works in a format-version-1 project with PROJECT_INDEX.md, docs/plans, and docs/decisions. Do not use for a small contained task that needs no durable plan, or when the user explicitly opts out of Scoville Plan.
scoville-ui-anti-ai-slop
Framework-aware guardrail for UI design, implementation, and audit. Use for hierarchy, layout, states, responsiveness, accessibility, usability, or clarity. Preserve the product design system and platform language. Excludes backend-only work and prose. Compose with Code for engineering and Scribe for variable UI wording.
compact-handoff
Create a copy-ready English Markdown prompt that transfers active work to another AI agent or session. Use when the user asks for a "compact handoff", "context handoff", "session handoff", "handoff to a new session", "Übergabe an neue Session", or otherwise explicitly requests a continuation prompt for another agent or session. Requires an explicit user request to hand work over. Do not use for requests to shorten, summarize, wrap up, or reduce context, and never activate on your own initiative because context is running low or a session is ending.
scoville-scribe-anti-ai-slop
Guardrail for prose, fidelity, and reader-facing interface text. Use for drafting, editing, summarizing, localizing, source-exact work, or audits. Preserve facts, meaning, terms, behavior, schemas, attribution, and exact text. Excludes code semantics, machine data, and supplied-text insertion without a wording, fidelity, or interface judgment.
ask-claude-for-codex
Ask Claude Code from Codex for a read-only second opinion, review, critique, comparison, or alternative analysis through the local Claude CLI. Use when the user says "Frage Fable", "ask Claude", "ask Opus", requests a Claude, Fable, or Opus opinion, wants another model to inspect the current workspace, or invokes $ask-claude-for-codex. Accept a Claude model alias or full model ID and an effort level; default to Fable 5 with high effort.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.